CSS/NTRES 6200

Spatial Modelling and Analysis

Raster GIS

D G Rossiter

Cornell UniversityDepartment of Crop & Soil Sciences

December 28, 2018

Copyright © 2018–19 Cornell University

All rights reserved. Reproduction and dissemination of the work as a whole (not parts) freely permitted if this

original copyright notice is included. Sale or placement on a web site where payment must be made to access this

document is strictly prohibited. To adapt or translate please contact the author

(http://www.css.cornell.edu/faculty/dgr2/).

Raster GIS 1

GIS data models

These are how spatial data are represented inside a GIS – not their conceptualrepresentation

Vector exact mathematical form: 0-dimension = points; 1-dimension = linesegments (which can be joined); 2-dimension = areas; 3-dimension = volumes

• Note: a Triangulated Irregular Network (TIN) is a vector data model of a 2-Dcontinuous surface conceptual model

Raster a regular tessellation (e.g., square or hexagonal grid) of a area

• 1D: line segments; 2D: “pixels” (from ‘picture elements’); 3D “voxels”(’volume pixels’)

• continuous-valued raster can be considered as a discretization of aspatially-continuous field

• e.g., remotely-sensed image

D G Rossiter

Raster GIS 2

Example raster DEM

CRS: EPSG:26718 - NAD27 / UTM zone 18N10 x 10 m horizontal resolution.dem: USGS Digital Elevation Model (DEM) data file

D G Rossiter

Raster GIS 3

Special characteristics of rasters

• fundamental difference with vector representation: fixed maximumresolution, fixed geometry

– can have compressed storage, e.g., quadtrees, but conceptually oneresolution (the finest)

• one data value per cell; (1) continuous feature-space variable

– single values at the centre point (typical for DEM)– grid cell (block) averages– direct sensor output (e.g., remote sensing), depends on point-spread

function– maxima, minima, other statistical summaries of within-cell information . . .

• one data value per cell; (2) classified feature-space variable

– the attribute, i.e., class name– a code linking to an attribute table with categories or continuous values– but it is more common to have a raster stack

D G Rossiter

Raster GIS 4

Support

• Support: physical size of data item

• Be very clear on what the value represents; support may not equal grid size

– larger support: one pixel in a map unit (e.g., gridded version of a vector soilmap – see gSSURGO)

– smaller support: "point" observation within the pixel used as its value– kriging: punctual (to grid centre) vs. block average; over= or ! = grid cell

size

• Conversion to/from point vector representation

– change of support? (not if raster value is the punctual value at the cellcentre)

D G Rossiter

Raster GIS 5

Example of larger support than pixel size

NRCS National Soil Survey Center, and NRCS National Geospatial Center of Excellence. 2014. Gridded Soil Survey

Geographic (gSSURGO) Database User Guide, Version 1.1. NRCS.

http://www.nrcs.usda.gov/wps/PA_NRCSConsumption/download?cid=nrcs142p2_051847&ext=pdf

D G Rossiter

Raster GIS 6

Example of support same as pixel size

Chaney, N. W. et al. (2016). POLARIS: A 30-meter probabilistic soil series map of the contiguous United States.

Geoderma, 274, 54–67. https://doi.org/10.1016/j.geoderma.2016.03.025

D G Rossiter

Raster GIS 7

Operations

• overlay: very fast

• area, border, length calculations very simple –> FRAGSTATS

• categorization of operations:

– local: at a pixel∗ e.g., transformations, map algebra

– focal: around a pixel, in its neighbourhood∗ e.g., filter

– zonal: pixel in some map unit or ‘zone’∗ e.g., mean value of all pixels in the map unit

– global: all pixels∗ e.g., distance from a source

D G Rossiter

Raster GIS 8

(source: Li Xingong, Univ. of Kansas)D G Rossiter

Raster GIS 9

Change of resolution

• upscaling by aggregation, various ways

– central pixel– average, mode– also make uncertainty layer of σ or entropy

• downscaling continuous-valued rasters by image fusion with finer-resolutionimages, downscaling cokriging etc.

• downscaling categorical-valued rasters by finer-scale information anddisaggregation models

D G Rossiter

Raster GIS 10

Example of disaggregation

original disaggregated

rasterized polygon map same legend, more spatial detail

source: Yang, L., et al.. 2011. Updating Conventional Soil Maps through Digital Soil Mapping. Soil Sci. Soc. Am. J.

75(3): 1044–1053.

D G Rossiter

Raster GIS 11

Finer-scale layers used in the disaggregation model

(a) slope gradient;(b) plan curvature;(c) profile curvature;(d) TWIAll these from fine-scale DEM

D G Rossiter

Raster GIS 12

Adjacency, neighbours

• Adjacency, neighbour concepts very simple

• Moran’s I weights can only be centroid distances or steps (queen vs. rook)

D G Rossiter